@7@fedilist@jeffcliff@shibao Prometheus, probably; CSV, yes. (R probably has a thing in CRAN for asking Prometheus questions. This is probably easier than working with Prometheus's web UI. I should have thought of that sooner.)
> an rss feed would be orders of magnitude more verbose
So there's also a CSV. :senko:
> prometheus endpoints are just a couple lines of whatever text you want to give at a certain point...
Yeah, but it's ugly. ( https://screamshitter.club/rvl/prometheus ) Semantic markup is better (parseable by selectors, presentable to the person using the browser as-is, with links and everything), there are plans for that. CSVs are better (very easy to use by piping them into awk), those are already present, RSS is better (wide variety of clients that will show you just new stuff). There are a million easier ways to get a single number into a TSDB, so there aren't any plans for Prometheus; if someone plans to actually use it, plans might form, but I don't have a use for it.
@shibao@jeffcliff@fedilist Yeah, I kinda hate them, though. I implemented some Prometheus stuff for Revolver but it is not helpful. I'd rather dump RSS and plain text and whoever can do whatever with that information.
I think the better thing would be to start tracking when things swing in a weird direction: number of posts or new accounts or new peers goes up faster than expected. This would be useful for people that want to spot anomalies (like the *.gab.best or *.activitypub-troll.cf instances creating massive spikes in peer lists for some servers, or a spambot creating a big spike in posts) or people that want to see when something interesting is going on with the network (e.g., which instances are starting to get more activity).
I'm trying to avoid doing anything to it that takes more than 15 or 20 minutes (because busy :liquidocelot:) so maybe instead of writing it, I'll just expose a firehose RSS feed and hope someone else writes it. All the changes, as they come in, as fast as anyone's bot cares to refresh, then they can do the data analysis.